Spectra and Median Size Mods

Size Analysis of the Northeast US Finfish Community

Published

April 26, 2024

Modeling Change in Median Size and Community Spectra

The following tabs detail how I’m thinking about whether to use SST in Celsius or as anomalies to test our hypotheses about temperature’s impact on size and size structure:

SST Change Over Time

Landings Change Over Time

Landings Verification

Bart raised that his landings looked different, vdrify that we didn’t mess them up in data prep:

Landings and Anomalies Colinearity

Landings Clustering/Bi-modality

Spectra Slope Models

Contemporary SST/Landings Model

effect	group	term	estimate	std.error	statistic
fixed	NA	(Intercept)	-0.921360717	0.009839746	-93.6366360
fixed	NA	landings	0.018576979	0.010705595	1.7352589
fixed	NA	regionGeorges_Bank	-0.012148242	0.012621505	-0.9625034
fixed	NA	regionMid_Atlantic_Bight	-0.114884622	0.012637417	-9.0908308
fixed	NA	regionSouthern_New_England	-0.117012827	0.012378032	-9.4532658
fixed	NA	sst_anom	-0.004065130	0.009320448	-0.4361518
fixed	NA	landings:regionGeorges_Bank	-0.006458479	0.012945264	-0.4989067
fixed	NA	landings:regionMid_Atlantic_Bight	-0.031802414	0.013679938	-2.3247484
fixed	NA	landings:regionSouthern_New_England	-0.005036802	0.012402315	-0.4061179
ran_pars	year	sd__(Intercept)	0.022544228	NA	NA
ran_pars	Residual	sd__Observation	0.050777216	NA	NA

Significance

	npar	AIC	LRT	Pr(Chi)
	NA	-436.8383	NA	NA
sst_anom	1	-438.6385	0.1998047	0.6548785
landings:region	3	-435.7297	7.1086049	0.0685156

Fit & Predictions

Collinearity

SST Effect

Landings Effect

5-Year Smoothed Predictor Model

effect	group	term	estimate	std.error	statistic
fixed	NA	(Intercept)	-0.925286340	0.009426778	-98.15509603
fixed	NA	land_5	0.019084432	0.009964401	1.91526141
fixed	NA	regionGeorges_Bank	-0.008202934	0.011842074	-0.69269404
fixed	NA	regionMid_Atlantic_Bight	-0.111854877	0.011990018	-9.32899978
fixed	NA	regionSouthern_New_England	-0.116485618	0.011704724	-9.95201773
fixed	NA	anom_5	-0.001404379	0.014197197	-0.09891946
fixed	NA	land_5:regionGeorges_Bank	0.005290628	0.012412550	0.42623218
fixed	NA	land_5:regionMid_Atlantic_Bight	-0.033785136	0.013692523	-2.46741497
fixed	NA	land_5:regionSouthern_New_England	-0.011822140	0.012133137	-0.97436792
ran_pars	year	sd__(Intercept)	0.022792284	NA	NA
ran_pars	Residual	sd__Observation	0.050168060	NA	NA

Significance

	npar	AIC	LRT	Pr(Chi)
	NA	-442.7303	NA	NA
anom_5	1	-444.7201	0.0101711	0.9196679
land_5:region	3	-439.2276	9.5026805	0.0233029

Fit & Predictions

Collinearity

SST Effect

Landings Effect

Median Body Weight Models

How has median weight changed

Contemporary SST/Landings Model

effect	group	term	estimate	std.error	statistic
fixed	NA	(Intercept)	0.176728059	0.009359083	18.883053
fixed	NA	landings	0.021297703	0.010155283	2.097204
fixed	NA	regionGeorges_Bank	-0.019398093	0.011926868	-1.626420
fixed	NA	regionMid_Atlantic_Bight	-0.124714909	0.011948834	-10.437413
fixed	NA	regionSouthern_New_England	-0.121706242	0.011696896	-10.405003
fixed	NA	sst_anom	-0.002809302	0.008914145	-0.315151
fixed	NA	landings:regionGeorges_Bank	-0.013659094	0.012239620	-1.115974
fixed	NA	landings:regionMid_Atlantic_Bight	-0.020025091	0.012965115	-1.544536
fixed	NA	landings:regionSouthern_New_England	-0.018056535	0.011724685	-1.540044
ran_pars	year	sd__(Intercept)	0.022163153	NA	NA
ran_pars	Residual	sd__Observation	0.047978561	NA	NA

Fit & Predictions

Significance

	npar	AIC	LRT	Pr(Chi)
	NA	-452.6072	NA	NA
sst_anom	1	-454.5014	0.1058008	0.7449765
landings:region	3	-455.4081	3.1991481	0.3619278

Collinearity

SST Effect

Landings Effect

5-Year Smoothed Predictor Model

effect	group	term	estimate	std.error	statistic
fixed	NA	(Intercept)	0.17088659	0.008870858	19.2638172
fixed	NA	land_5	0.02683093	0.009353268	2.8686152
fixed	NA	regionGeorges_Bank	-0.01466959	0.011082561	-1.3236639
fixed	NA	regionMid_Atlantic_Bight	-0.12022181	0.011224651	-10.7105166
fixed	NA	regionSouthern_New_England	-0.11868733	0.010954052	-10.8350156
fixed	NA	anom_5	0.01107564	0.013407216	0.8260954
fixed	NA	land_5:regionGeorges_Bank	-0.00506618	0.011619881	-0.4359924
fixed	NA	land_5:regionMid_Atlantic_Bight	-0.02535527	0.012852623	-1.9727704
fixed	NA	land_5:regionSouthern_New_England	-0.02201549	0.011357699	-1.9383757
ran_pars	year	sd__(Intercept)	0.02197317	NA	NA
ran_pars	Residual	sd__Observation	0.04694883	NA	NA

Fit & Predictions

Significance

	npar	AIC	LRT	Pr(Chi)
	NA	-461.8534	NA	NA
anom_5	1	-463.1299	0.7234317	0.3950205
land_5:region	3	-461.3420	6.5113491	0.0892160

Collinearity

SST Effect

Landings Effect

Region Random Effects

What if we swap Random effects

I don’t really care about regions, so what if we used region as a random effect. It should also eat variability similar to year as a random effect and maybe help resolve the temperature or landings effects.

Just to see:

Contemporary SST/Landings Model

effect	group	term	estimate	std.error	df	statistic	p.value
fixed	NA	(Intercept)	0.109362822	0.028243777	145	3.8721033	0.0001627259
fixed	NA	landings	0.004371285	0.005859928	145	0.7459624	0.4568977897
fixed	NA	sst_anom	-0.003793204	0.008033209	145	-0.4721904	0.6375007698
ran_pars	survey_area	sd_(Intercept)	0.054752813	NA	NA	NA	NA
ran_pars	Residual	sd_Observation	0.053207811	NA	NA	NA	NA

Fit & Predictions

Significance

	Df	AIC	LRT	Pr(>Chi)
	NA	-440.8509	NA	NA
landings	1	-442.2872	0.5636696	0.4527855
sst_anom	1	-442.6271	0.2237871	0.6361694

Collinearity

SST Effect

Landings Effect

5-Year Smoothed Predictor Model

effect	group	term	estimate	std.error	statistic
fixed	NA	(Intercept)	-0.98269973	0.031874118	-30.8306483
fixed	NA	land_5	0.00173992	0.006155699	0.2826518
fixed	NA	anom_5	0.01589537	0.016403943	0.9689969
fixed	NA	land_5:anom_5	0.02932610	0.013546341	2.1648727
ran_pars	survey_area	sd__(Intercept)	0.06289653	NA	NA
ran_pars	Residual	sd__Observation	0.05544279	NA	NA

Significance

	npar	AIC	LRT	Pr(Chi)
	NA	-424.4753	NA	NA
land_5:anom_5	1	-421.7740	4.701356	0.0301388

Fit & Predictions

Collinearity

SST Effect

Landings Effect

Variance Partitioning

--- title: "Spectra and Median Size Mods" description: | Size Analysis of the Northeast US Finfish Community date: "`r Sys.Date()`" format: html: self-contained: true code-fold: true code-tools: true df-print: kable toc: true toc-depth: 2 execute: echo: false warning: false message: false fig.align: center comment: "" --- # Modeling Change in Median Size and Community Spectra ```{r} library(gt) library(tidyverse) library(gmRi) library(tidyverse) library(ggeffects) library(zoo) library(lme4) library(patchwork) library(emmeans) library(rstatix) library(lmerTest) library(broom) library(broom.mixed) library(ggpmisc) library(performance) library(nlme) # Set theme theme_set(theme_gmri() + theme(plot.margin = margin(), legend.position = "bottom", legend.direction = "horizontal") ) # levels for faceting areas area_levels <- c("Northeast Shelf", "GoM", "GB", "SNE", "MAB") area_levels_long <- c("Northeast Shelf", "Gulf of Maine", "Georges Bank", "Southern New England", "Mid-Atlantic Bight") #-------------------- ## Set up #-------------------- # Load the raw data df <- read_csv(here::here("data/size_and_spectra_model_data.csv")) # Get raw landings landings_raw <- read_csv(here::here("data/unscaled_spectra_predictor_dataset_wide.csv")) %>% select(year, survey_area, landings_raw = landings) # Get the raw sst from oisst sst_raw = read_csv(here::here("data/unscaled_regional_sst.csv")) #------------ # Put it all together and clarify the column names df <- df %>% select(-sst) %>% left_join(landings_raw) %>% left_join(sst_raw) %>% mutate( survey_area = factor( survey_area, levels = area_levels_long), region = str_replace_all(survey_area, "-| ", "_"), region = factor( region, levels = c("Gulf_of_Maine", "Georges_Bank", "Mid_Atlantic_Bight", "Southern_New_England"))) # format year and anomalies df <- df %>% mutate(yr_num = as.numeric(as.character(year)), year = factor(year)) %>% group_by(survey_area) %>% arrange(survey_area, year) %>% mutate( anom_5 = zoo::rollapply( sst_anom, 5, mean, na.rm = T, align = "right", fill = NA), land_5_raw = zoo::rollapply( landings_raw, 5, mean, na.rm = T, align = "right", fill = NA)) ``` The following tabs detail how I'm thinking about whether to use SST in Celsius or as anomalies to test our hypotheses about temperature's impact on size and size structure: ### SST Change Over Time ```{r} # Plot: How does temperature change over time sst_1 <- df %>% mutate(year = as.numeric(as.character(year))) %>% drop_na(sst) %>% ggplot(aes(year, sst, color = region)) + geom_point(size = 1, alpha = 0.6) + geom_smooth(method = "lm", linewidth = 1, se = F) + scale_color_gmri() + labs(title = "Regional SST Increasing") # Plot: Anomaly Changes sst_2 <- df %>% mutate(year = as.numeric(as.character(year))) %>% drop_na(sst_anom) %>% ggplot(aes(year, sst_anom, color = region)) + geom_point(size = 1, alpha = 0.6) + geom_smooth(method = "lm", linewidth = 1, se = F) + scale_color_gmri() + labs(title = "Regional SST Anomalies Increasing") sst_1 / sst_2 ``` ### Landings Change Over Time ```{r} # Plot: How does landings change over time df %>% mutate(year = as.numeric(as.character(year))) %>% drop_na(landings_raw) %>% ggplot(aes(year, landings_raw, color = region)) + geom_point() + geom_smooth(method = "lm") + scale_color_gmri() + scale_y_continuous(labels = scales::label_comma()) + labs(title = "Regional Landings Declining", subtitle = "Mid-Atlantic Exception, primarily forage fish fishery (menhaden)") # Trying to keep landings: binned/categorical label ``` ### Landings Verification Bart raised that his landings looked different, vdrify that we didn't mess them up in data prep: ```{r} #| label: garfo-data-prep #| eval: true #### GARFO Landings #### # Load the GARFO landings data: # Landings of finfish* sheet 5 res_path <- cs_path("res") landings <- readxl::read_xlsx( path = str_c(res_path, "GARFO_landings/KMills_landings by area 1964-2021_JUN 2022.xlsx"), sheet = 5) %>% rename_all(tolower) # # Why no GB Landings in 2011? # landings %>% filter( # `stat area` %in% fish_zones$"Georges Bank", year == 2011) # Make a list of zones to roughly match the survey areas: fish_zones <- list( "Gulf of Maine" = c(511:515, 464, 465), "Georges Bank" = c(521, 522, 525, 561, 562), "Southern New England" = c(611, 612, 613, 616, 526, 537, 538, 539), "Mid-Atlantic Bight" = c(614:615, 621, 622, 625, 626, 631, 632)) # Add the labels into the landings data and remove what we don't need there: landings <- landings %>% mutate( survey_area = case_when( `stat area` %in% fish_zones$"Gulf of Maine" ~ "Gulf of Maine", `stat area` %in% fish_zones$"Georges Bank" ~ "Georges Bank", `stat area` %in% fish_zones$"Southern New England" ~ "Southern New England", `stat area` %in% fish_zones$"Mid-Atlantic Bight" ~ "Mid-Atlantic Bight")) %>% filter(survey_area %in% c("Georges Bank", "Gulf of Maine", "Southern New England", "Mid-Atlantic Bight")) %>% mutate(survey_area = factor(survey_area, area_levels_long)) # Get Summaries landings_summ <- landings %>% rename( "weight_lb" = `landed lbs`, "live_lb" = `live lbs`) %>% group_by(year, survey_area) %>% summarise( across( .cols = c(value, weight_lb, live_lb), .fns = list(mean = ~mean(.x , na.rm = T), total = ~sum(.x , na.rm = T)), .names = "{.fn}_{.col}"), .groups = "drop") # Scale the landings to create an index by area landings_summ <- landings_summ %>% group_by(survey_area) %>% mutate(total_wt_z = as.numeric(base::scale(total_weight_lb))) %>% ungroup() landings_summ %>% filter(year >= 1970) %>% ggplot(aes(year, total_weight_lb, color = survey_area)) + geom_point() + geom_smooth(method = "lm") + scale_color_gmri() + scale_y_continuous(labels = scales::label_comma()) + labs(title = "Regional Landings Declining", subtitle = "Mid-Atlantic Exception, primarily forage fish fishery (menhaden)") ``` ### Landings and Anomalies Colinearity ```{r} # See roll anom and roll landings relationship ggplot(df, aes(anom_5, land_5_raw)) + geom_point() + facet_wrap(~region) + stat_poly_eq() # cor.test(df$sst_anom, df$landings_raw) #-.24 # cor.test(df$anom_5, df$land_5_raw) # -.35 # cor.test(df$sst_anom, df$landings) # -.46 # cor.test(df$anom_5, df$land_5) # -.6 ``` ### Landings Clustering/Bi-modality ```{r} # Exploring bimodality of landings ggplot(df, aes(landings)) + geom_density(aes(color = survey_area)) ``` ```{r} #| label: landings-clustering #| eval: false # clustering landings # for 2 clusters region_mods <- df %>% drop_na(landings) %>% split(.$region) %>% imap(function(x,y){ mod = kmeans(drop_na(x,landings)$landings, centers = 2) data.frame( region = y, year = x$year, cluster = mod$cluster )}) # use betweens to assess which value for centers maximizes between group SS # evaluate n-clusters clust_range <- c(2:9) clust_num_mods <- clust_range %>% map(function(x){ landings_df <- drop_na(df,landings) mod = kmeans(landings_df$landings, centers = x) list( "cluster_dat" = data.frame( year = landings_df$year, region = landings_df$region, cluster = mod$cluster), "betweens" = tibble("Between Group SS" = mod$betweens)) }) # Pull performance map_dfr(set_names(clust_num_mods, clust_range), ~pluck(.x, "betweens"), .id = "num_clusters") %>% ggplot(aes(num_clusters, `Between Group SS`)) + geom_point() + geom_line() # Pull the cluster info, make sure that each region has good representation df %>% ggplot(aes(landings)) + geom_density() + facet_wrap(~region) + labs(title = "Landings (scaled) PDF") ``` --- # Spectra Slope Models ## Contemporary SST/Landings Model ```{r} ## 5-year smooth on anomalies ## temperature impacts on growth and/or migration at large scales not instantaneous and integrate exposure over long time periods ## # new model # From low to high correlation between landings and sst anomalies # Landings in raw form should be scaled # new_mod <- lme4::lmer(b ~ landings_raw * region + sst_anom + (1 | year), data = df) # new_mod <- lme4::lmer(b ~ land_5_raw * region + anom_5 + (1 | year), data = df) # To smooth or not to smooth, I vote no new_mod <- lme4::lmer(b ~ landings * region + sst_anom + (1 | year), data = drop_na(df, landings, sst_anom)) # new_mod <- lme4::lmer(b ~ land_5 * region + anom_5 + (1 | year), data = df) # new_mod <- lme4::lmer(b ~ landings * region * sst_anom + (1 | year), data = df) # # Autocorrelative Error - Performs badly here # Ben bolker to thatnk, of course: # https://bbolker.github.io/mixedmodels-misc/notes/corr_braindump.html # new_mod <- nlme::lme(b ~ landings * region + sst_anom, # random = ~ 1 | year, # data = drop_na(df, landings, sst_anom), # correlation=corAR1(), method = "ML") # Summary tidy(new_mod) %>% gt() # # % variance of random effect # # year (intercept) / year (intercept) + residual # 0.0005082/(0.0005082 + 0.0025783) ``` ### Significance ```{r} drop1(new_mod, test = "Chisq") ``` ### Fit & Predictions ```{r} qqnorm(resid(new_mod)) qqline(resid(new_mod)) ggplot(df, aes(x = yr_num, y = b, color = region)) + geom_point(alpha = 0.5) + theme_classic() + scale_color_gmri() + geom_line( data = cbind(drop_na(df, landings, sst_anom), pred = predict(new_mod)), aes(y = pred), size = 1) + # adding predicted line from mixed model theme(legend.position = "none", panel.spacing = unit(2, "lines")) # adding space between panels ``` ### Collinearity ```{r} # multicollinearity check col_check <- check_collinearity(new_mod) plot(col_check) ``` ### SST Effect ```{r} # SST Anomaliezs plot(ggpredict(new_mod, ~sst_anom+region), add.data = T, jitter = 0.1) ``` ### Landings Effect ```{r} plot(ggpredict(new_mod, ~landings+region), add.data = T, jitter = 0.1) ``` ## 5-Year Smoothed Predictor Model ```{r} # Both sst & landings - clears collinearity test lag_mod_full <- lme4::lmer(b ~ land_5 * region * anom_5 + (1 | year), data = drop_na(df, land_5, anom_5)) lag_mod <- lme4::lmer(b ~ land_5 * region + anom_5 + (1 | year), data = drop_na(df, land_5, anom_5)) lag_mod_sst <- lme4::lmer(b ~ anom_5 * region + (1 | year), data = drop_na(df, land_5, anom_5)) lag_mod_landings <- lme4::lmer(b ~ land_5 * region + (1 | year), data = drop_na(df, land_5, anom_5)) # Summary tidy(lag_mod) %>% gt() ``` ### Significance ```{r} # drop1(lag_mod_full, test = "Chisq") drop1(lag_mod, test = "Chisq") # drop1(lag_mod_sst, test = "Chisq") # drop1(lag_mod_landings, test = "Chisq") ``` ### Fit & Predictions ```{r} qqnorm(resid(lag_mod)) qqline(resid(lag_mod)) ggplot(df, aes(x = yr_num, y = b, color = region)) + geom_point(alpha = 0.5) + theme_classic() + scale_color_gmri() + geom_line( data = cbind(drop_na(df, land_5, anom_5), pred = predict(lag_mod)), aes(y = pred), size = 1) + # adding predicted line from mixed model theme(legend.position = "none", panel.spacing = unit(2, "lines")) # adding space between panels ``` ### Collinearity ```{r} # multicollinearity check # col_check <- check_collinearity(lag_mod_full) col_check <- check_collinearity(lag_mod) # col_check <- check_collinearity(lag_mod_landings) # col_check <- check_collinearity(lag_mod_sst) plot(col_check) ``` ### SST Effect ```{r} # SST Anomaliezs plot(ggpredict(lag_mod, ~anom_5~region), add.data = T, jitter = 0.1) ``` ### Landings Effect ```{r} plot(ggpredict(lag_mod, ~land_5+region), add.data = T, jitter = 0.1) ``` --- # Median Body Weight Models How has median weight changed ## Contemporary SST/Landings Model ```{r} # new model new_mod <- lme4::lmer(med_wt ~ landings * region + sst_anom + (1 | year), data = drop_na(df, landings, sst_anom)) # Summary tidy(new_mod) %>% gt() ``` ### Fit & Predictions ```{r} qqnorm(resid(new_mod)) qqline(resid(new_mod)) ggplot(df, aes(x = yr_num, y = med_wt, color = region)) + geom_point(alpha = 0.5) + theme_classic() + scale_color_gmri() + geom_line( data = cbind(drop_na(df, landings, sst_anom), pred = predict(new_mod)), aes(y = pred), size = 1) + # adding predicted line from mixed model theme(legend.position = "none", panel.spacing = unit(2, "lines")) # adding space between panels ``` ### Significance ```{r} drop1(new_mod, test = "Chisq") ``` ### Collinearity ```{r} # multicollinearity check col_check <- check_collinearity(new_mod) plot(col_check) ``` ### SST Effect ```{r} # SST Anomaliezs plot(ggpredict(new_mod, ~sst_anom+region), add.data = T, jitter = 0.1) ``` ### Landings Effect ```{r} plot(ggpredict(new_mod, ~landings+region), add.data = T, jitter = 0.1) ``` ## 5-Year Smoothed Predictor Model ```{r} lag_mod <- lme4::lmer(med_wt ~ land_5 * region + anom_5 + (1 | year), data = drop_na(df, land_5, anom_5)) # Summary tidy(lag_mod) %>% gt() ``` ### Fit & Predictions ```{r} qqnorm(resid(lag_mod)) qqline(resid(lag_mod)) ggplot(df, aes(x = yr_num, y = med_wt, color = region)) + geom_point(alpha = 0.5) + theme_classic() + scale_color_gmri() + geom_line( data = cbind(drop_na(df, land_5, anom_5), pred = predict(lag_mod)), aes(y = pred), size = 1) + # adding predicted line from mixed model theme(legend.position = "none", panel.spacing = unit(2, "lines")) # adding space between panels ``` ### Significance ```{r} drop1(lag_mod, test = "Chisq") ``` ### Collinearity ```{r} # multicollinearity check col_check <- check_collinearity(lag_mod) plot(col_check) ``` ### SST Effect ```{r} # SST Anomaliezs plot(ggpredict(lag_mod, ~anom_5~region), add.data = T, jitter = 0.1) ``` ### Landings Effect ```{r} plot(ggpredict(lag_mod, ~land_5+region), add.data = T, jitter = 0.1) ``` --- # Region Random Effects # What if we swap Random effects I don't really care about regions, so what if we used region as a random effect. It should also eat variability similar to year as a random effect and maybe help resolve the temperature or landings effects. Just to see: ## Contemporary SST/Landings Model ```{r} # new model new_mod <- lme4::lmer(med_wt ~ landings * sst_anom + (1 | survey_area), data = drop_na(df, landings, sst_anom)) # # Autocorrelative Error - Performs badly here # Ben bolker to thatnk, of course: # https://bbolker.github.io/mixedmodels-misc/notes/corr_braindump.html new_mod <- nlme::lme(med_wt ~ landings * landings + sst_anom, random = ~ 1 | survey_area, data = drop_na(df, landings, sst_anom), correlation = corAR1(), method = "ML") # Summary tidy(new_mod) %>% gt() ``` ### Fit & Predictions ```{r} qqnorm(resid(new_mod)) qqline(resid(new_mod)) ggplot(df, aes(x = yr_num, y = med_wt, color = region)) + geom_point(alpha = 0.5) + theme_classic() + scale_color_gmri() + geom_line( data = cbind(drop_na(df, landings, sst_anom), pred = predict(new_mod)), aes(y = pred), size = 1) + # adding predicted line from mixed model theme(legend.position = "none", panel.spacing = unit(2, "lines")) # adding space between panels ``` ### Significance ```{r} drop1(new_mod, test = "Chisq") ``` ### Collinearity ```{r} # multicollinearity check col_check <- check_collinearity(new_mod) plot(col_check) ``` ### SST Effect ```{r} # SST Anomaliezs plot(ggpredict(new_mod, ~sst_anom), add.data = T, jitter = 0.1) ``` ### Landings Effect ```{r} plot(ggpredict(new_mod, ~landings), add.data = T, jitter = 0.1) ``` ## 5-Year Smoothed Predictor Model ```{r} # Both sst & landings lag_mod <- lme4::lmer(b ~ land_5 * anom_5 + (1 | survey_area), data = drop_na(df, land_5, anom_5)) # Summary tidy(lag_mod) %>% gt() ``` ### Significance ```{r} drop1(lag_mod, test = "Chisq") ``` ### Fit & Predictions ```{r} qqnorm(resid(lag_mod)) qqline(resid(lag_mod)) ggplot(df, aes(x = yr_num, y = b, color = region)) + geom_point(alpha = 0.5) + theme_classic() + scale_color_gmri() + geom_line( data = cbind(drop_na(df, land_5, anom_5), pred = predict(lag_mod)), aes(y = pred), size = 1) + # adding predicted line from mixed model theme(legend.position = "none", panel.spacing = unit(2, "lines")) # adding space between panels ``` ### Collinearity ```{r} # multicollinearity check col_check <- check_collinearity(lag_mod) plot(col_check) ``` ### SST Effect ```{r} # SST Anomaliezs plot(ggpredict(lag_mod, ~anom_5), add.data = T, jitter = 0.1) ``` ### Landings Effect ```{r} plot(ggpredict(lag_mod, ~land_5), add.data = T, jitter = 0.1) ``` # Variance Partitioning ```{r} library(vegan) ```